Prediction of Air Quality Index (AQI) Based on Optimized GWO-LSTM Neural Network

With the booming economy and the massive emission of polluting gases from factories, the air quality problems are seriously affecting human's physical and mental health; and effective air quality prediction methods would play an early warning role for pollutants. This paper proposes an improved gray wolf algorithm (GWO) optimized long short-term memory network (LSTM) method to predict Air Quality Index (AQI). In order to overcome the shortcomings of traditional LSTM network using back propagation, that is easy to fall into local optimum, the connection weights of LSTM are updated iteratively by improving the gray wolf algorithm; the Levy flight strategy is used to guide the head wolf to find the optimum, and only head wolf
$a$
is taken to guide the update of wolf pack position to ensure the global search ability of the algorithm within the early stage and the local exploitation ability in the later stage to improve the prediction accuracy of the LSTM network. Finally, we do comparison experiments on the improved LSTM and the prediction of AQI using different models.